Future projections from CNRM-CM6 model (6 variables, best algo from ffs)

1 Context

One GCM at a 25 km resolution were tested (“CNRM-CM6”), including Sea Surface Temperature (tos), Sea Surface Salinity (sos), Sea Surface Height (zos) and Mixed Layer Depth (mlotst). These 4 dynamic variables were first (i) extracted from CMIP6 for the historical period (1993-2014) and contemporary period (2015-2020) from ssp126. Then, each variable was included in the initial models run on the available dataset (current predictions: “present”).

The results here are presented for summer only.

2 Model performance

2.1 Bowheads

Figure 2.1: Boxplot of the performance metrics for bowheads in summer.

Strong inter-run variability for bowheads East with earth algorithm (MARS), so the best algo chosen was the blackboost (BRT) for both sides; but still quite low performance east side (especially low specificity: does not predict well whale’s presence).

2.2 Narwhals

Figure 2.2: Boxplot of the performance metrics for narwhals in summer.

The Random Forest suffers from overfitting, so the best algorithm is the NNET for narwhals East and blackboost (BRT) for narwhals West.

2.3 Belugas

Figure 2.3: Boxplot of the performance metrics for belugas in summer.

The best algo is the NNET for belugas.

3 Belugas

3.1 Present

Figure 3.1: Max prediction maps for belugas in present summer from the NNET.

3.2 Mean 2030

Figure 3.2: Mean projection maps for Belugas in summer 2030 from the NNET (ssp126).

Figure 3.3: Mean projection maps for Belugas in summer 2030 from the NNET (ssp585).

3.3 Mean 2060

Figure 3.4: Mean projection maps for Belugas in summer 2060 from the NNET (ssp126).

Figure 3.5: Mean projection maps for Belugas in summer 2060 from the NNET (ssp585).

3.4 Mean 2100

Figure 3.6: Mean projection maps for Belugas in summer 2100 from the NNET (ssp126).

Figure 3.7: Mean projection maps for Belugas in summer 2100 from the NNET (ssp585).

4 Narwhals

4.1 Present

Figure 4.1: Mean prediction maps for Narwhals in present summer from the NNEt (East) and BRT (West).

4.2 Mean 2030

Figure 4.2: Mean projection maps for Narwhals in summer 2030 from the BRT and NNET (ssp585).

Figure 4.3: Mean projection maps for Narwhals in summer 2030 from the BRT and NNET (ssp585).

4.3 Mean 2060

Figure 4.4: Mean projection maps for Narwhals in summer 2060 from the BRT and NNET (ssp585).

Figure 4.5: Mean projection maps for Narwhals in summer 2060 from the BRT and NNET (ssp585).

4.4 Mean 2100

Figure 4.6: Mean projection maps for Narwhals in summer 2100 from the BRT and NNET (ssp585).

Figure 4.7: Mean projection maps for Narwhals in summer 2100 from the BRT and NNET (ssp585).

5 Bowheads

5.1 Present (BRT)

Figure 5.1: Mean prediction maps for Bowheads in present summer from the BRT and MARS.

5.2 Mean 2030

Figure 5.2: Mean projection maps for Bowheads in summer 2030 from the BRT (ssp126).

Figure 5.3: Mean projection maps for Bowheads in summer 2030 from the BRT (ssp585).

5.3 Mean 2100

Figure 5.4: Mean projection maps for Bowheads in summer 2100 from the BRT (ssp126).

Figure 5.5: Mean projection maps for Bowheads in summer 2100 from the BRT (ssp585).